Conditional disease‐free survival rates and their associated determinants in patients with epithelial ovarian cancer: A 15‐year retrospective cohort study

Abstract Background The most common type of ovarian cancer (OC) is epithelial ovarian cancer (EOC) which is the most lethal gynecologic malignancy in adult women. Aim This study aimed to determine the conditional disease‐free survival (CDFS) rates and their associated determinants in patients with EOC. Methods and results The clinical and demographic data of 335 patients with confirmed EOC at Motahari Clinic (Shiraz, Iran) were retrospectively reviewed and analyzed. Traditional DFS (TDFS) and CDFS were calculated using the Kaplan–Meier method and cumulative DFS estimates, respectively. To evaluate the effects of the prognostic determinants on the DFS of the patients, a multiple covariate Cox analysis using the landmarking method was applied. The 1‐ and 3‐year TDFSs were 81.1% and 47.0%, respectively, and decreased over time. At baseline, a higher stage tumor and endometrioid histology were associated with a higher risk of recurrence when compared to stage I and other histological subtypes, respectively. The hazard of recurrence for older women (age ≥55 years) was approximately twice and three times more than that of women aged <45 years at 1‐ and 3‐year landmark time points, respectively. Conclusion The age at diagnosis, defined by a cut‐off of 55 years, was a prognostic factor for the CDFS of EOC women. Moreover, patients with advanced‐stage EOC (ASEOC) (stages III and IV) and endometrioid histology had poorer CDFSs compared to those with early‐stage EOC (ESEOC) (stages I and II) and other histological types. In ESEOC patients with age at diagnosis of >55 years, CDFS gradually decreased in 3 years after remission which should be considered for follow‐up care decision‐making.


| INTRODUCTION
Ovarian cancer (OC) is considered the seventh most common cancer among women worldwide. The most common type of OC is epithelial ovarian cancer (EOC) which is the most lethal gynecologic malignancy in adult women. 1 The estimated overall 5-year survival rate for earlystage epithelial ovarian cancer patients (ESEOC: stages I and II) and patients with advanced-stage epithelial ovarian cancer (ASEOC: stages III and IV) varies between 55% and 92% and 18% and 30%, respectively. 2,3 The statistical measures made at the time of diagnosis are traditionally used for evaluating the survival rate. In this regard, two types of traditionally used survival estimates in previous studies are overall survival (OS) and disease-free survival (DFS) which are reported from the time of diagnosis and remission, respectively. [3][4][5][6] Despite providing important information for both clinicians and patients, such analyses appear to be of less value for patients who survive a period of time after their initial diagnosis and treatment. 7 Although most patients with EOC attain remission, most of them eventually relapse. Follow-up attentions including the close monitoring of CA-125 levels, imaging tests, and physical exams are necessary and important in affecting disease outcomes. 8,9 A more accurate estimate for these patients is conditional DFS (CDFS) which considers changes in the patients' recurrence risks over time. 7 Moreover, a recent clinical trial demonstrated that initiating chemotherapy for EOC patients with CA-125 elevation and without symptoms did not offer any more benefits than delaying treatment until the emergence of clinical evidence of disease recurrence. 10 Additionally, the early treatment of the relapsed EOC patients based on the rise of CA-125 levels alone has resulted in the earlier deterioration of their quality of life. 10 In these situations, using traditional DFS (TDFS) may not provide realistic and optimistic information about the recurrence risk for EOC patients, and CDFS estimates might offer more accurate information regarding the risk of recurrence so that the best-informed decisions about the patients' follow-up care can be made. 7 Today, the risk of recurrence from the time of treatment or surgery can be computed using the available instruments for predicting the recurrence risk of EOC. CDFS estimates are based on the definition of conditional survival probability. These estimates incorporate the dynamic change in the survival risk over time. Therefore, they have been recognized as a more meaningful indicator of the survival probability of patients who have an initial survival period. 7,11 The median DFS was reported to be 2.54 years (range: 0.03-9.96 years) among patients with OC in a previous study. Moreover, 3-year DFS was 48.2%. 12 The stage and histological type can affect the OS of patients with EOC. Moreover, age ≥ 60 years was correlated with poorer overall survival than age <60 years in a previous study. 13 Another study showed that chemotherapeutic regimens/cycles and tumor grade and stage were independent prognostic factors for early-stage EOC.
Moreover, no significant differences were observed between 5-year OS and histological types. 14 Considering the insufficient data in the literature, this study aimed to determine the conditional disease-free survival (CDFS) rates and their associated determinants in Iranian patients with EOC.  Patients with nonepithelial ovarian cancer, tumors   with an unknown or not applicable stage, borderline histology tumor, undefined histological types, without pathology confirmation, and lost to follow-up were excluded. Besides, the patients who did not provide signed informed consent for the analysis of their medical records and those with nonanalyzable data (deficit in data) were also excluded.
The age at diagnosis, marriage status, child-bearing (parity), tumor histology, and the FIGO (International Federation of Gynecology and Obstetrics) stage were considered for the analysis. The FIGO classification was used for disease staging and histological grading. 6 Stages I and II were considered as ESEOC and stages III and IV were regarded as ASEOC. As the first-line chemotherapy regimen, platinum-paclitaxel was given to patients eligible for chemotherapy (except for patients with stage IA and grade I disease).
According to the opinion of the physician and the tolerance of patients to the side effects of chemotherapy, three to six cycles of chemotherapy were applied.

| Analysis of the survival time
The primary endpoint was DFS. The time interval between the initial remission date and the recurrence date or the last contact was considered as the DFS time. Patients without diagnosed recurrent EOC during the follow-up period were censored at the end of the study. The date of diagnosis was defined as the date of the primary surgery in patients without cytology or the date of the first positive cytology.
The date on which no evidence of disease was found by the oncologist was considered as the date of remission. If this information was missing, the date of the first negative surgical or imaging result or the date of the first factor indicating no evidence of disease such as normal levels of CA-125 was used for defining the date of remission.
When none of these criteria were available, 1 month after the date of chemotherapy completion or (if no chemotherapy was given) the date of the primary surgery was used. Recurrence was defined as a return of clinical symptoms on follow-up after the patient has been in remission for a while. The date of recurrence was defined based on a process similar to the one applied for the date of remission. If the date on which an oncologist first diagnosed any signs of recurrence was available, it was considered as the date of recurrence. The date of the first positive surgical or imaging result, the initiation date of chemotherapy/radiation, or the date of the first evidence indicating disease recurrence such as elevated CA-125 levels after being disease-free for a while was used as the recurrence date. The TDFS rates were calculated by nonparametric Kaplan-Meier (NPKM) estimates and compared using the log-rank test. 15 The simultaneous impact of various patients' characteristics on the DFS was analyzed using the multiple covariate Cox-adjusted proportional hazards (PH) model and employing the Breslow method for the ties. After assessing the PH assumption (using the goodness-of-fit testing approach based on Schoenfeld residuals), the findings were interpreted using hazard ratios (HRs). The confidence intervals (CIs) for the HRs were computed based on Wald's test of the Cox-adjusted PH regression parameters. The landmark analysis method was applied with six landmark time points (at baseline and after 1, 2, 3, 4, and 5 years from the baseline).

| Conditional disease-free survival (CDFS)
CDFS estimates were directly calculated from traditional NPKM disease-free survival estimates. CDFS is defined as the probability of staying disease-free for an additional number of years (t 2 ) provided that a patient has already been in remission for t 1 number of years. It is expressed as CDFS (t 2 jt 1 ). The exponential version of Greenwood's formula (see Appendix A) was used to compute the CDFS estimates with a confidence interval (CI) of 95%. 16 The changes in DFS over time were evaluated by comparing the estimations of 1-and 3-year CDFSs (CDFS (1jx) and CDFS (3jx)) at 1, 2, 3, 4, and 5 years after attaining remission with baseline 1-and 3-year DFS estimates. In addition to the overall CDFS, to assess the impact of EOC women's characteristics, 1-and 3-year CDFS estimates were also calculated within the strata defined by age at diagnosis, marriage status, parity, stage, and histology.
All the statistical analyses were performed using the R software (version: 3.6.2) and GraphPad Prism software (version: 6.07). A p-value of ≤.05 was considered statistically significant.

| RESULTS
After applying the exclusion criteria to a total of 600 EOC patients who were evaluated, 335 patients were considered for the analysis.
The median (95% CI) OS was 3.58 (3.00-4.17) years (ranging from 0.25 to 13.33 years). About 61.2% of the women were diagnosed with recurrent EOC, and 38.8% of them were still alive at the termination of the study. The median (95% CI) DFS for the women was 2.75 14.6% of the study population had been diagnosed with stages I, II, III, and IV EOC, respectively. 53.0%, 6.7%, 32.9%, and 7.4% of the women who survived 5 years without recurrence had stages I, II, III, and IV EOC, respectively. Moreover, 81.8%, 10.7%, 3.3%, and 4.2% of the patients were diagnosed with serous epithelial, mucinous, and endometrioid carcinomas as well as other types of tumor (i.e., Brenner, undifferentiated carcinoma, and clear cell carcinoma) at baseline, respectively, while the EOC women who survived 5 years without recurrence were 76.5%, 15.4%, 3.4%, and 4.7% in the abovementioned histological subgroups, respectively.
Using the nonparametric log-rank test analysis, the age of >55 years at diagnosis led to lower TDFS estimates (p = .0001) and women with ESEOC had significantly higher TDFS estimates than those with ASEOC (p < .0001). Moreover, the TDFS curves were not statistically significant in terms of marriage status (p = .1454), different histological subtypes (p = .1107), and child-bearing or parity (0.0901) (Figure 1).
Landmark analyses were applied based on the multiple covariate Cox-adjusted PH regression (Table 1)  At baseline, the 1-and 3-year DFSs were 81.1% and 47.0%, respectively, and decreased over time. The 1-and 3-year CDFS estimates for EOC women are depicted in Figure 2. Based on the concept of CDFS, the probabilities of staying disease-free for an additional 1 year were 73%, 79%, 86%, 88%, and 90%, respectively, given that the patient has already been in remission for x years (i.e., CDFS (1jx)) and x = 1, 2, 3, 4, and 5. Moreover, the probability of surviving an additional 3 years without recurrence, conditioned on having already survived x years after remission (i.e., CDFS (3jx)) when x = 1, 2, 3, 4, and 5 improved to 50%, 60%, 68%, 64%, and 66%, respectively   Figure 3, the probability of staying disease-free for an additional 3 years given that an EOC woman with mucinous carcinoma has already been in remission for 4 years is equal to 0.61, that is, CDFS (3j4) = 0.61. In contrast to serous and mucinous histology subtypes, the 3-year CDFS of the patients with endometrioid carcinoma gradually decreased over time (see Figure 3B). The hazard ratios (95% CIs) for recurrence using landmark analysis based on the multiple covariate Cox-adjusted regression at baseline and the 1st, 2nd, 3rd, 4th, and 5th years of remission    A previous study found initial differences in the DFS of OC patients at the time of remission among stage, age at diagnosis, grade, and histology groups which diminished over time. 12 However, the differences in 3-year CDFS estimates based on age at diagnosis, histology, and stage remained high and did not diminish over time in the present study. Therefore, it can be inferred that each of these characteristics should be considered separately for each patient. Regarding   26 However, a previous study demonstrated that the mortality and morbidity percentages were the same across younger and older patients who were equally debulked. 27 Another study also found that age was not an independent prognostic factor for either OS or DFS among EOC women and performance status affected the treatment outcome for elderly patients. 22 In general, there is still controversy about the prognostic effect of age in EOC women.
The subgroup analysis based on histology indicated that the 3-year CDFS of the EOC women with endometrioid carcinoma gradually decreased during the 3 years after remission. This result was in contrast to the 3-year CDFS of patients with serous carcinoma and mucinous carcinoma ( Figure 4B). Based on the results of the present study, endometrioid carcinoma has the worst prognosis among the histological types of EOC. Therefore, patients with endometrioid carcinoma need regular and more intense follow-up examinations.
Previous research also found an improvement in the OS rate for EOC women with clear cell and endometrioid carcinomas and mucinous cystadenocarcinoma. 18 The OS of stage IV EOC women with mucinous and clear cell subtypes was significantly lower than that of women with other histological subtypes in another previous work as well. 28 However, due to the few numbers of endometrioid subtypes in the present study (11 cases out of 335), it was not possible to compare the results of this study with those of previous studies directly.
A study conducted on EOC patients indicated that more than one parity improved OS in ASEOC patients. 3 However, the results of the current study did not show any relationship between the CDFS of EOC women and their parity.
One of the goals of follow-up is the early detection of disease recurrence. However, there is no consensus on whether increased surveillance for recurrent disease can necessarily improve OS. 18,29 Besides, no clear guidelines are available on the type and frequency of follow-up care. One of the earliest indicators of disease recurrence in EOC patients is a more than twice rise of the upper limit of normal CA-125. However, the early treatment of tumor recurrence based on the increased level of CA-125 did not demonstrate a significant survival benefit. 10 The results of landmark analyses showed that age at diagnosis and histological type were predictive of subsequent DFS after 1 to 5 years of remission. These were in line with the results of a previous study which found that the main prognostic factor for the DFS of the EOC patients was tumor histology [28]. Moreover, it has been reported that age at diagnosis, the results of primary surgery and second-look surgery, stage of disease, optimal debulking surgery,

ETHICS STATEMENT
The study protocol was approved by the Ethics Committee for Research of Shiraz University of Medical Sciences (Protocol# IR. SUMS.REC.1393.8910) and was conducted in compliance with the Declaration of Helsinki. All patients had given written informed consent to allow the use of their personal and clinical data retrospectively.

PATIENT CONSENT
All study participants, or their legal guardian, provided written informed consent prior to study enrollment. Besides, the patients who did not provide signed informed consent for the analysis of their medical records were excluded.

DATA AVAILABILITY STATEMENT
The data used to support the findings of the present study are available from the corresponding author upon request.